Automatic Chinese Catchword Extraction Based on Time Series Analysis

CoNLL 2008: Proceedings of the 12th Conference on Computational Natural Language Learning, pages 113–118 Manchester, August 2008 Automatic Chinese Catchword Extraction Based on Time Series Analysis Han Ren1, Donghong Ji1, Jing Wan2 and Lei Han1 1 School of Computer Science, Wuhan University 430079, China 2 Center for Study of Language & Information, Wuhan University 430072, China cslotus@mail.whu.edu.cn, donghong_ji@yahoo.com, jennifer.wanj@gmail.com, hattason@mail.whu.edu.cn Abstract Catchwords refer to those popular words or phrases in a time period. In this paper, we propose a novel approach for automatic extraction of Chinese catchwords. By analyzing features of catchwords, we define three aspects to describe Popular Degree of catchwords. Then we use curve fitting in Time Series Analysis to build Popular Degree Curves of the extracted terms. Finally we give a formula that can calculate Popular Degree values of catchwords and get a ranking list of catchword candidates. Experiments show that the method is effective. 1 Introduction Generally, a catchword is a term which represents a hot social phenomenon or an important incident, and is paid attention by public society within certain time period. On the one hand, catchwords represent the mass value orientation for a period. On the other hand, they have a high timeliness. Currently, there are quiet a few ranking and evaluations of catchwords every year in various kinds of media. Only in year 2005, tens of Chinese organizations published their ranking list of Chinese catchwords. Catchwords contain a great deal of information from any particular area, and such words truly and vividly reflect changes of our lives and our society. By monitoring and analysis of catchwords, we can learn the change of public © 2008. Licensed under the Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported license (http://creativecommons.org/licenses/by-nc- sa/3.0/). Some rights reserved. attention in time. In addition, we may detect the potential changes of some linguistic rules, which can help establish and adjust state language policies. Currently, two kinds of approaches are adopted to evaluate catchwords. One is by CTR (Click-Through Rate) or retrieval times, but the limitation is that it is just based on frequency, which is only one feature of catchwords. The other is by manual evaluation, but it depends on their subjective judgment to a large extent. In this paper, we propose a novel approach that can automatically analyze and extract Chinese catchwords. By analyzing sample catchwords and finding out their common features, we provide a method to evaluate the popular degree. After ranking, terms that have high values are picked out as catchword candidates. The rest of the paper is organized as follows. In Section 2, we discuss about the linguistic basis of catchword judgment. In Section 3, we describe the extraction method in detail. In Section 4, we present the experimental results as well as some discussions. Finally, we give the conclusion and future work in Section 5. 2 Linguistic basis The popularity of a word or phrase contains two factors: time and area, namely how long it lasts and how far it spreads. But neither of them have definite …

Content truncated to 500 words